Performance evaluation for four classes of textural features

Abstract Textural features for pattern recognition are compared. The problem addressed is to determine which features optimize classification rate. Such features may be used in image segmentation, compression, inspection, and other problems in computer vision. Many textural features have been proposed in the literature. No large-scale objective comparative study has appeared. The goal is comparing and evaluating in a quantitative manner four types of features, namely Markov Random Field parameters, multi-channel filtering features, fractal based features, and co-occurrence features. Performance is assessed by the criterion of classification error rate with a Nearest Neighbor classifier and the Leave-One-Out estimation method using forward selection. Four types of texture are studied, two synthetic (fractal and Gaussian Markov Random Fields) and two natural (leather and painted surfaces). The results show that co-occurrence features perform best followed by the fractal features. However, there is no universally best subset of features. The feature selection task has to be performed for each specific problem to decide which feature of which type one should use.